Am I understanding correctly that this would only prevent scheduling new backups from myhoard's internal scheduling loop, not external requests? That seems like a feature, but perhaps good to emphasize here.

What happens if the existing backups get stuck for some reason? Does that mean we just never schedule anything new - presumably we'd still get a backupage alert in that case?

I would perhaps wait and see how the cloud provider fixes progress since at the moment it looks like we cannot rely on the object storage API and it may not be meaningful to try to work around the most critical component the software deals with being broken.

Closing for now in favour of investigation into how to handle unstable object storage APIs as the real root cause.